A comparative study of multilayered and single layered neural network based predictive controllers

The authors compare the performance of two types of controllers one based on the multilayered network and the other based on the single layered CMAC network (cerebellar model articulator controller). The neurons (information processing units) in the multi-layered network use Gaussian activation functions. The control scheme which is considered is a predictive control algorithm, along the lines used by Willis et al. (1991), Kambhampati and Warwick (1991). The process selected as a test bed is a continuous stirred tank reactor. The reaction taking place is an irreversible exothermic reaction in a constant volume reactor cooled by a single coolant stream. This reactor is a simplified version of the first tank in the two tank system given by Henson and Seborg (1989).

Evaluation of satellite-based and model re-analysis rainfall estimates for Uganda

The dependence of much of Africa on rain fed agriculture leads to a high vulnerability to fluctuations in rainfall amount. Hence, accurate monitoring of near-real time rainfall is particularly useful, for example in forewarning possible crop shortfalls in drought-prone areas. Unfortunately, ground based observations are often inadequate. Rainfall estimates from satellite-based algorithms and numerical model outputs can fill this data gap, however rigorous assessment of such estimates is required. In this case, three satellite based products (NOAA-RFE 2.0, GPCP-1DD and TAMSAT) and two numerical model outputs (ERA-40 and ERA-Interim) have been evaluated for Uganda in East Africa using a network of 27 rain gauges. The study focuses on the years 2001 to 2005 and considers the main rainy season (February to June). All data sets were converted to the same temporal and spatial scales. Kriging was used for the spatial interpolation of the gauge data. All three satellite products showed similar characteristics and had a high level of skill that exceeded both model outputs. ERA-Interim had a tendency to overestimate whilst ERA-40 consistently underestimated the Ugandan rainfall.

Validation of the FALL3D ash dispersion model using observations of the 2010 Eyjafjallajökull volcanic ash clouds

During April-May 2010 volcanic ash clouds from the Icelandic Eyjafjallajökull volcano reached Europe causing an unprecedented disruption of the EUR/NAT region airspace. Civil aviation authorities banned all flight operations because of the threat posed by volcanic ash to modern turbine aircraft. New quantitative airborne ash mass concentration thresholds, still under discussion, were adopted for discerning regions contaminated by ash. This has implications for ash dispersal models routinely used to forecast the evolution of ash clouds. In this new context, quantitative model validation and assessment of the accuracies of current state-of-the-art models is of paramount importance. The passage of volcanic ash clouds over central Europe, a territory hosting a dense network of meteorological and air quality observatories, generated a quantity of observations unusual for volcanic clouds. From the ground, the cloud was observed by aerosol lidars, lidar ceilometers, sun photometers, other remote-sensing instru- ments and in-situ collectors. From the air, sondes and multiple aircraft measurements also took extremely valuable in-situ and remote-sensing measurements. These measurements constitute an excellent database for model validation. Here we validate the FALL3D ash dispersal model by comparing model results with ground and airplane-based measurements obtained during the initial 14e23 April 2010 Eyjafjallajökull explosive phase. We run the model at high spatial resolution using as input hourly- averaged observed heights of the eruption column and the total grain size distribution reconstructed from field observations. Model results are then compared against remote ground-based and in-situ aircraft-based measurements, including lidar ceilometers from the German Meteorological Service, aerosol lidars and sun photometers from EARLINET and AERONET networks, and flight missions of the German DLR Falcon aircraft. We find good quantitative agreement, with an error similar to the spread in the observations (however depending on the method used to estimate mass eruption rate) for both airborne and ground mass concentration. Such verification results help us understand and constrain the accuracy and reliability of ash transport models and it is of enormous relevance for designing future operational mitigation strategies at Volcanic Ash Advisory Centers.

The system identification and control of Hammerstein system using non-uniform rational B-spline neural network and particle swarm optimization

In this paper a new system identification algorithm is introduced for Hammerstein systems based on observational input/output data. The nonlinear static function in the Hammerstein system is modelled using a non-uniform rational B-spline (NURB) neural network. The proposed system identification algorithm for this NURB network based Hammerstein system consists of two successive stages. First the shaping parameters in NURB network are estimated using a particle swarm optimization (PSO) procedure. Then the remaining parameters are estimated by the method of the singular value decomposition (SVD). Numerical examples including a model based controller are utilized to demonstrate the efficacy of the proposed approach. The controller consists of computing the inverse of the nonlinear static function approximated by NURB network, followed by a linear pole assignment controller.

A model-based PID controller for Hammerstein systems using B-spline neural networks

In this paper, a new model-based proportional–integral–derivative (PID) tuning and controller approach is introduced for Hammerstein systems that are identified on the basis of the observational input/output data. The nonlinear static function in the Hammerstein system is modelled using a B-spline neural network. The control signal is composed of a PID controller, together with a correction term. Both the parameters in the PID controller and the correction term are optimized on the basis of minimizing the multistep ahead prediction errors. In order to update the control signal, the multistep ahead predictions of the Hammerstein system based on B-spline neural networks and the associated Jacobian matrix are calculated using the de Boor algorithms, including both the functional and derivative recursions. Numerical examples are utilized to demonstrate the efficacy of the proposed approaches.

Modular non-computational-connectionist-hybrid neural network approach to robotic systems

Spiking neural networks are usually limited in their applications due to their complex mathematical models and the lack of intuitive learning algorithms. In this paper, a simpler, novel neural network derived from a leaky integrate and fire neuron model, the ‘cavalcade’ neuron, is presented. A simulation for the neural network has been developed and two basic learning algorithms implemented within the environment. These algorithms successfully learn some basic temporal and instantaneous problems. Inspiration for neural network structures from these experiments are then taken and applied to process sensor information so as to successfully control a mobile robot.

A thermoresponsive supramolecular polymer network comprising pyrene-functionalized gold nano-particles and a chain-folding polydiimide.

A thermoresponsive, supramolecular nanocomposite has been prepared by the addition of pyrenyl functionalized gold nanoparticles (AuNPs) to a polydiimide that contains receptor residues designed to form defined complexes with pyrene. The novel pyrenyl-functionalized AuNPs (P-AuNPs) were characterized by transmission electron microscopy, with surface functionalization confirmed by infrared and UV–visible spectroscopic analyses. Mixing solutions of the P-AuNPs and a π-electron-deficient polydiimide resulted in the formation of electronically complementary, chain-folded and π–π-stacked complexes, so affording a new supramolecular nanocomposite network which precipitated from solution. The P-AuNPs bind to the polydiimide via π–π stacking interactions to create supramolecular cross-links. UV–visible spectroscopic analysis confirmed the thermally reversible nature of the complexation process, and transmission electron microscopy (TEM), infrared spectroscopy (IR), and differential scanning calorimetry (DSC) were used to characterize the supramolecular-nanocomposite material. The supramolecular polymer network is insoluble at room temperature, yet may be dissolved at temperatures above 60 °C. The thermal reversibility of this system is maintained over five heat/cool cycles without diminishment of the network characteristics. In contrast to the individual components, the nanocomposite formed self-supporting films, demonstrating the benefit of the supramolecular network in terms of mechanical properties. Control experiments probing the interactions between a model diimide compound that can also form a π-stacked complex with the π-electron rich pyrene units on P-AuNPs showed that, while complexation was readily apparent, precipitation did not occur because a supramolecular cross-linked network system could not be formed with this system.

Adaptive noise cancellation with fast tunable RBF network

This paper describes a novel adaptive noise cancellation system with fast tunable radial basis function (RBF). The weight coefficients of the RBF network are adapted by the multi-innovation recursive least square (MRLS) algorithm. If the RBF network performs poorly despite of the weight adaptation, an insignificant node with little contribution to the overall performance is replaced with a new node without changing the model size. Otherwise, the RBF network structure remains unchanged and only the weight vector is adapted. The simulation results show that the proposed approach can well cancel the noise in both stationary and nonstationary ANC systems.

Online modeling with tunable RBF network

In this paper, we propose a novel online modeling algorithm for nonlinear and nonstationary systems using a radial basis function (RBF) neural network with a fixed number of hidden nodes. Each of the RBF basis functions has a tunable center vector and an adjustable diagonal covariance matrix. A multi-innovation recursive least square (MRLS) algorithm is applied to update the weights of RBF online, while the modeling performance is monitored. When the modeling residual of the RBF network becomes large in spite of the weight adaptation, a node identified as insignificant is replaced with a new node, for which the tunable center vector and diagonal covariance matrix are optimized using the quantum particle swarm optimization (QPSO) algorithm. The major contribution is to combine the MRLS weight adaptation and QPSO node structure optimization in an innovative way so that it can track well the local characteristic in the nonstationary system with a very sparse model. Simulation results show that the proposed algorithm has significantly better performance than existing approaches.

Lunar and solar FTIR nitric acid measurements at Eureka in winter 2001/2002: comparisons with observations at Thule and Kiruna and with CMAM and SLIMCAT model calculations

For the first time, vertical column measurements of (HNO3) above the Arctic Stratospheric Ozone Observatory (AStrO) at Eureka (80N, 86W), Canada, have been made during polar night using lunar spectra recorded with a Fourier Transform Infrared (FTIR) spectrometer, from October 2001 to March 2002. AStrO is part of the primary Arctic station of the Network for the Detection of Stratospheric Change (NDSC). These measurements were compared with FTIR measurements at two other NDSC Arctic sites: Thule, Greenland (76.5N, 68.8W) and Kiruna, Sweden (67.8N, 20.4E). The measurements were also compared with two atmospheric models: the Canadian Middle Atmosphere Model (CMAM) and SLIMCAT. This is the first time that CMAM HNO3 columns have been compared with observations in the Arctic. Eureka lunar measurements are in good agreement with solar ones made with the same instrument. Eureka and Thule HNO3 columns are consistent within measurement error. Differences among HNO3 columns measured at Kiruna and those measured at Eureka and Thule can be explained on the basis of the available sunlight hours and the polar vortex location. The comparison of CMAM HNO3 columns with Eureka and Kiruna data shows good agreement, considering CMAM small inter-annual variability. The warm 2001/02 winter with almost no Polar Stratospheric Clouds (PSCs) makes the comparison of the warm climate version of CMAM with these observations a good test for CMAM under no PSC conditions. SLIMCAT captures the magnitude of HNO3 columns at Eureka, and the day-to-day variability, but generally reports higher HNO3 columns than the CMAM climatological mean columns.

Sensitivity analysis of an ocean carbon cycle model in the North Atlantic: an investigation of parameters affecting the air-sea CO2 flux, primary production and export of detritus

The sensitivity of the biological parameters in a nutrient-phytoplankton-zooplankton-detritus (NPZD) model in the calculation of the air-sea CO2 flux, primary production and detrital export is analysed. We explore the effect on these outputs of variation in the values of the twenty parameters that control ocean ecosystem growth in a 1-D formulation of the UK Met Office HadOCC NPZD model used in GCMs. We use and compare the results from one-at-a-time and all-at-a-time perturbations performed at three sites in the EuroSITES European Ocean Observatory Network: the Central Irminger Sea (60° N 40° W), the Porcupine Abyssal Plain (49° N 16° W) and the European Station for Time series in the Ocean Canary Islands (29° N 15° W). Reasonable changes to the values of key parameters are shown to have a large effect on the calculation of the air-sea CO2 flux, primary production, and export of biological detritus to the deep ocean. Changes in the values of key parameters have a greater effect in more productive regions than in less productive areas. The most sensitive parameters are generally found to be those controlling well-established ocean ecosystem parameterisations widely used in many NPZD-type models. The air-sea CO2 flux is most influenced by variation in the parameters that control phytoplankton growth, detrital sinking and carbonate production by phytoplankton (the rain ratio). Primary production is most sensitive to the parameters that define the shape of the photosynthesis-irradiance curve. Export production is most sensitive to the parameters that control the rate of detrital sinking and the remineralisation of detritus.

Estimates of global multicomponent aerosol optical depth and direct radiative perturbation in the Laboratoire de Météorologie Dynamique general circulation model

The global cycle of multicomponent aerosols including sulfate, black carbon (BC),organic matter (OM), mineral dust, and sea salt is simulated in the Laboratoire de Me´te´orologie Dynamique general circulation model (LMDZT GCM). The seasonal open biomass burning emissions for simulation years 2000–2001 are scaled from climatological emissions in proportion to satellite detected fire counts. The emissions of dust and sea salt are parameterized online in the model. The comparison of model-predicted monthly mean aerosol optical depth (AOD) at 500 nm with Aerosol Robotic Network (AERONET) shows good agreement with a correlation coefficient of 0.57(N = 1324) and 76% of data points falling within a factor of 2 deviation. The correlation coefficient for daily mean values drops to 0.49 (N = 23,680). The absorption AOD (ta at 670 nm) estimated in the model is poorly correlated with measurements (r = 0.27, N = 349). It is biased low by 24% as compared to AERONET. The model reproduces the prominent features in the monthly mean AOD retrievals from Moderate Resolution Imaging Spectroradiometer (MODIS). The agreement between the model and MODIS is better over source and outflow regions (i.e., within a factor of 2).There is an underestimation of the model by up to a factor of 3 to 5 over some remote oceans. The largest contribution to global annual average AOD (0.12 at 550 nm) is from sulfate (0.043 or 35%), followed by sea salt (0.027 or 23%), dust (0.026 or 22%),OM (0.021 or 17%), and BC (0.004 or 3%). The atmospheric aerosol absorption is predominantly contributed by BC and is about 3% of the total AOD. The globally and annually averaged shortwave (SW) direct aerosol radiative perturbation (DARP) in clear-sky conditions is �2.17 Wm�2 and is about a factor of 2 larger than in all-sky conditions (�1.04 Wm�2). The net DARP (SW + LW) by all aerosols is �1.46 and �0.59 Wm�2 in clear- and all-sky conditions, respectively. Use of realistic, less absorbing in SW, optical properties for dust results in negative forcing over the dust-dominated regions.

Does the PMSP connectionist model of single word reading learn to read in the same way as a child?

The Plaut, McClelland, Seidenberg and Patterson (1996) connectionist model of reading was evaluated at two points early in its training against reading data collected from British children on two occasions during their first year of literacy instruction. First, the network’s non-word reading was poor relative to word reading when compared with the children. Second, the network made more non-lexical than lexical errors, the opposite pattern to the children. Three adaptations were made to the training of the network to bring it closer to the learning environment of a child: an incremental training regime was adopted; the network was trained on grapheme– phoneme correspondences; and a training corpus based on words found in children’s early reading materials was used. The modifications caused a sharp improvement in non-word reading, relative to word reading, resulting in a near perfect match to the children’s data on this measure. The modified network, however, continued to make predominantly non-lexical errors, although evidence from a small-scale implementation of the full triangle framework suggests that this limitation stems from the lack of a semantic pathway. Taken together, these results suggest that, when properly trained, connectionist models of word reading can offer insights into key aspects of reading development in children.

The real-time optimisation of DNO owned storage devices on the LV network for peak reduction

Energy storage is a potential alternative to conventional network reinforcementof the low voltage (LV) distribution network to ensure the grid’s infrastructure remainswithin its operating constraints. This paper presents a study on the control of such storagedevices, owned by distribution network operators. A deterministic model predictive control (MPC) controller and a stochastic receding horizon controller (SRHC) are presented, wherethe objective is to achieve the greatest peak reduction in demand, for a given storagedevice specification, taking into account the high level of uncertainty in the prediction of LV demand. The algorithms presented in this paper are compared to a standard set-pointcontroller and bench marked against a control algorithm with a perfect forecast. A specificcase study, using storage on the LV network, is presented, and the results of each algorithmare compared. A comprehensive analysis is then carried out simulating a large number of LV networks of varying numbers of households. The results show that the performance of each algorithm is dependent on the number of aggregated households. However, on a typical aggregation, the novel SRHC algorithm presented in this paper is shown to outperform each of the comparable storage control techniques.

A radial basis function network classifier to maximise leave-one-out mutual information

tWe develop an orthogonal forward selection (OFS) approach to construct radial basis function (RBF)network classifiers for two-class problems. Our approach integrates several concepts in probabilisticmodelling, including cross validation, mutual information and Bayesian hyperparameter fitting. At eachstage of the OFS procedure, one model term is selected by maximising the leave-one-out mutual infor-mation (LOOMI) between the classifier’s predicted class labels and the true class labels. We derive theformula of LOOMI within the OFS framework so that the LOOMI can be evaluated efficiently for modelterm selection. Furthermore, a Bayesian procedure of hyperparameter fitting is also integrated into theeach stage of the OFS to infer the l2-norm based local regularisation parameter from the data. Since eachforward stage is effectively fitting of a one-variable model, this task is very fast. The classifier construc-tion procedure is automatically terminated without the need of using additional stopping criterion toyield very sparse RBF classifiers with excellent classification generalisation performance, which is par-ticular useful for the noisy data sets with highly overlapping class distribution. A number of benchmarkexamples are employed to demonstrate the effectiveness of our proposed approach.